| In this thesis, image coding using multirate filter banks and wavelet systems are evaluated and compared. The coding systems considered are based on the M-channel general lapped biorthgonal transforms (GLBT) and the wavelet-based subband coding schemes, as well as the baseline JPEG algorithm. The principles behind each algorithm are briefly described. The thesis concentrates on both coding efficiency and complexity. The tradeoff between efficiency and complexity of each coding system has also been analyzed.; From the analysis of lossy image coding results, we find that the GLBT algorithm provides a compromise solution between the JPEG algorithm and the wavelet-based scheme due to its characteristics of moderate efficiency and complexity. Therefore, the GLBT algorithm is an alternative solution to current image coding standards. On the other hand, JPEG is the fastest algorithm and therefore presumably the least complex. The performance of several wavelet-based transforms is also evaluated on the same metrics. It reveals that the SPIHT algorithm which is the basis of the JPEG-2000 standard provides the most effective and flexible solution. Through the implementation of our software package, which is based on the EZW algorithm, it is worth noting that the main factors in image coding are the quantizer and entropy coder rather than the transforms only. The high performance of the wavelet algorithm is offset by its computing complexity. The coding results also show that the choice of an image coding scheme for a particular application depends strongly on its requirements. |
